Content uploaded by Luca Bonini
Author content
All content in this area was uploaded by Luca Bonini on Apr 13, 2019
Content may be subject to copyright.
Fabrication and Characterization of a High-Resolution Neural Probe
for Stereoelectroencephalography and Single Neuron Recording
F. Pothof, S. Anees, J. Leupold, L. Bonini, O. Paul, Member IEEE, G.A. Orban, P. Ruther, Member IEEE
Abstract— This paper reports on the design, fabrication, and
characterization of neural probes for stereoelectroencephalo-
graphy (SEEG). The probe specifically targets focal epilepsy as
key application. However, probes of this type can also be used
for the diagnosis and treatment of other neural dysfunctions
such as Parkinson’s disease or tremor, typically requiring deep
brain probes. The probe fabrication relies on a novel fabrication
concept for rolling and gluing thin film polyimide sheets with
integrated electrodes into permanent cylindrical shapes with
diameters down to 800 µm. The SEEG probes comprise several
macro-electrodes designed to record local field potentials, and
micro-electrodes positioned in-between, dedicated to monitoring
single unit activity. While platinum micro-electrodes with a
diameter of 35 µm have impedances of about 245 kΩat
1 kHz, impedance values down to about 2 kΩhave been
measured for the macro-electrodes. The devices have shown
good compatibility with magnetic resonance imaging in a 9.4 T
magnet, enabling the precise post-operative probe localization
within the brain.
I. INTRODUCTION
Depth probes with multiple electrodes aligned along the
probe axis are used in various clinical applications, such as
the treatment of Parkinson’s disease based on deep brain
stimulation (DBS), and the diagnosis of focal epilepsy. In
this later case, stereoelectroencephalography (SEEG), intro-
duced by Talairach and Bancaud in the 1950s [1], exploits
depth probes to precisely localize the epileptogenic zone via
triangulation in patients with drug resistant focal epilepsy
in order to surgically resect the involved brain area [2].
Depth probes for these clinical applications most often use
platinum-iridium (PtIr) cylinders as electrodes aligned in a
pearl chain arrangement with insulating sections between
these recording sites. The individual electrodes are con-
nected to an external connector using single wires imposing
challenging space constraints in case of a large number of
electrodes. Depth probes for DBS typically have a diameter
of 1.27 mm and comprise four electrodes. On the other
hand, SEEG probes are available with diameters as small
as 0.8 mm drastically reducing the impaired brain volume
The research leading to these results has received funding from the
European Union’s Seventh Framework Program (FP7/2007-2013) under
grant agreement n◦600925.
F. Pothof, S. Anees, O. Paul, and P. Ruther are with the Microsystem
Materials Laboratory, Department of Microsystems Engineering (IMTEK),
University of Freiburg, Freiburg, Germany (phone +49-761-203-7200,
frederick.pothof@imtek.de).
J. Leupold is with the University Medical Center Freiburg, Department
of Radiology, Medical Physics, Freiburg, Germany.
G. Orban is with the Department of Neurosciences, Universit`
a degli studi
di Parma, Parma, Italy.
L. Bonini is with the Istituto Italiano di Tecnologia (IIT), Brain Center
for Social and Motor Cognition, Parma, Italy.
requested in view of up to 15 probes per hemisphere applied
in individual patients. These SEEG probes comprise up to 18
electrodes with a length of 2 mm, as demonstrated by Dixi
Medical, Besancon, France. Besides the cylindrical macro-
contacts of SEEG probes recording local field potentials
(LFP), additional micro-electrodes were integrated by Dixi
enabling single unit activity (SUA) to be recorded. This
approach has the potential to improve the spatial resolution
of triangulation in focal epilepsy diagnosis.
Recent neural probe developments have applied microsys-
tems fabrication processes using polymer layer sandwiches
with integrated electrodes. These planar structures have been
shaped into probe cylinders by adhesively bonding them to
solid carrier pins of appropriate diameters. Examples with
64 individually addressable electrodes have been demon-
strated [3] and alternative electrode configurations have been
described in a patent disclosure [4]. In contrast to these
examples, we introduce an approach similar to that used for
cuff electrode fabrication [5]. We realized initially hollow
cylinders with diameters as small as 800 µm. During fabrica-
tion, the cylinders are sealed using a structured dry adhesive
and stiffened by filling them with a medical grade epoxy
resin.
II. PROB E DESIGN
In a first design iteration of the novel SEEG probes
schematically shown in Fig. 1(a), the outer diameter was
chosen to be similar to that of existing probes for clinical
applications. As the probes are initially destined for in vivo
recordings with macaque monkeys for validation purposes,
the overall probe length was limited to 42 mm. Based on the
(a) (b)
Interface to zero
insertion force connector
Probe cylinder
Micro-
electrode
Macro-
electrode
Cytop
2 mm
2 mm
PI grid
Interface to zero
insertion force connector
Pro
be
cyl
ind
er
Micro-
ele
ctr
ode
Macro-
electrode
2 mm
gri
Fig. 1. (a) Schematic of SEEG probe geometry and (b) layout of a flat
polyimide layer sandwich foil, illustrating macro- and micro-contacts, later
brought into cylindrical shape through 3D profiling.
probe implantation protocol described in Section IV, 32 mm
of the total length can be implanted into brain tissue. The
probes designed with a diameter of 800 µm comprise in
total 16 or 32 channels, i.e., 8 cylindrical macro contacts
with a length of 2 mm each, and 8 or 24 micro-electrodes
with a diameter of 35 µm. The macro-electrodes comprise
a polyimide (PI) grid, as indicated in Fig. 1(b), to protect
the underlying platinum (Pt) from scratches or delamination
during the 3D profiling process. The macro-electrodes are
separated by a distance of 1.5 mm. The 8 or 24 micro-
electrodes are positioned around the circumference of the
cylinders between the macro-contacts. All electrodes are
interfaced to zero insertion force (ZIF) connectors. For the
ZIF mating process, a 0.3-mm-thick PCB is introduced in
parallel at the backside of the probe ZIF interface to guide
the flexible foil into the ridgid connector.
III. PROB E FABRICATION
The probes are based on a 10-µm-thick PI layer stack
comprising a 300-nm-thin Pt metallization for leads and
electrodes. To promote long-term adhesion between the PI
layers and the metallization sandwiched in-between, the Pt
leads are additionally covered by diamond like carbon (DLC)
and silicon carbide (SiC) layers [6]. The PI-metal sandwich,
as schematically shown in Fig. 1(b), is fabricated on a handle
wafer and transferred to the cylindrical probe shape using an
annealing step inside a metal mold. The fixation of the probe
in its cylindrical shape applies a dry resist that has been
applied during the fabrication of the PI layer sandwich. The
probe is finally filled with a medical grade epoxy to achieve
a probe stiffness appropriate for implantation. The probe
fabrication applies a two-stage process: First the PI layer
sandwich comprising the electrode metallization is realized;
second, by 3D-profiling of this layer sandwich cylinders with
a diameter of 800 µm are produced.
A. Stage 1 - Processing of PI Structure
Fig. 2 details the process sequences of the first stage. It is
based on a process technology introduced for the realization
of retinal implants [6]. In case of the SEEG probes developed
in this study, the process comprises additional steps to
(e) Spin coat Cytop
(f) Patterning PI
(f) Patterning Cytop
(d) Patterning Al
(a) Spin coat polyimide
(b) Patterning Pt, SiC + DLC
(c) Spin coat polyimide
Silicon PI LOR5A AZ1518
SiC/DLC Pt Al Cytop
Fig. 2. Probe fabrication based on thin film processing of PI, Pt, Al, SiC,
and DLC.
deposit and pattern a Cytop R
layer, i.e. a fluoropolymer by
Asahi Glass Company, Japan, on top of the PI. Similar to PI
ribbon cables used for silicon-based probe arrays developed
by our group [7][8], PI (U-Varnish S from UBE Industries
Ltd., Japan) is spin-coated with a layer thickness of 5 µm
onto standard four-inch silicon (Si) wafers {Fig. 2(a)}. In
order to facilitate the final removal of the PI layer sandwich
from the Si handle wafer, the native silicon oxide layer is
removed from the Si using an HF dip. PI spin-coating and
imidization at 450◦C is followed by the patterning of SiC
(50 nm), Pt (300 nm), SiC (30 nm), and DLC (10 nm)
using a lift-off technique based on LOR 5A and AZ 1518
photoresists. Pt and SiC as well as DLC are deposited using
sputtering and plasma enhanced chemical vapor deposition
(PECVD), respectively {Fig. 2(b)}. Before layer deposition,
an oxygen plasma treatment is used to activate the PI surface
improving the adhesion of the first SiC film deposited at
100◦C. This initial SiC film forms a thin interatomic interface
to the PI and thus promotes the long-term adhesion to the PI
layer [9]. Then, a second 5-µm-thick PI layer is spin-coated
on top of the SiC/Pt/SiC/DLC layer stack {Fig. 2(c)}.
Next, a 100-nm-thick aluminum (Al) layer is deposited and
patterned using sputtering and lift-off {Fig. 2(d)}. The Cytop
layer serving as a dry adhesive in the subsequent 3D-profiling
stage is spin coated on top of the PI sandwich and the Al
layer and then soft-baked {Fig. 2(e)}. It is patterned using
lithography and reactive ion etching (RIE) {Fig. 2(f)}. The
Cytop is removed on top of the Al layer that serves as a hard
mask to protect the underlying PI sandwich against the RIE
processing. The Al layer is finally removed by wet etching.
In a last process sequence, the PI is selectively etched by RIE
to open the electrode sites and contact pads, and to define the
probe shape {Fig. 2(g)}. This etching step removes possible
DLC or SiC residues on the Pt surface as well. The PI layer
stack is finally peeled off the wafer using tweezers.
B. Stage 2 - 3D Profiling of PI Sandwich
The 3D profiling of the SEEG probes is achieved through
rolling of the PI foils and annealing at elevated temperatures,
as schematically shown in Fig. 3(a), using a metallic mold
Cytop bonds to polyimide
800 µm
PolyimideCytop
Transfer to
mold
Hard bake:
290°C
(a)
(b) (c) (d)
Slit
Mold
PI-foil
Cytop
20 mm
Needle
Fig. 3. (a) Schematic of the rolling and bonding processes of polyimide
cylinders employing Cytop; (b) Computer model of the mold, (c) close-up
of the needle grabing the PI-foil, and (d) PI-foil rolled-up inside the cavity.
and slitted needle. The mold comprises 42-mm-long hollow
cylinders with an inner diameter of 800 µm accessible
from the mold top through narrow slits with a width of
approximately 80 µm. A model of the mold is shown in
Fig. 3(b). The hollow cylinders and slits are realized using
electrical discharge machining (EDM) of stainless steel.
The PI layer stacks are manually introduced into the mold
cylinders through the narrow slits. They are rolled into their
cylindrical shape using a slitted needle {Fig. 3(c)}(outer
diameter 700 µm; slit realized by EDM) rotated 2.5-times
inside the mold cylinder, as indicated in Fig. 3(d). The PI
foil and Cytop overlap by 800 µm. Exposing the rolled PI
layer stack inside the mold for 4 hours to a temperature
of 290◦C anneals the PI layer and hard bakes the Cytop
film. Due to the excellent adhesion of Cytop to PI, the PI
cylinder is fixed to its intended diameter using the Cytop
layer as a dry adhesive at the position of the overlapping
PI sections. Finally, the probe is filled with a biocompatible
epoxy (EPO-TEK R
301-2, Epoxy Technology, Inc., USA)
to ensure mechanical stability.
IV. APP LIC ATI ON PROTOCOL
The SEEG probes {Fig. 4(a)}are designed to be applied
in head-fixed macaque monkeys for validation before human
use. During their implantation, the probes with lengths
of several centimeters need to be guided to the targeted
brain area with sub-millimeter precision minimizing their
axial misalignment. This guidance will be accomplished by
custom-made hollow bone screws to which the probes will
be fixed. These screws are similar to those used in clinical
applications such as epilepsy diagnosis based on SEEG
probes by Dixi Medical. The screws for monkey and human
application are compared in Fig. 4(b). In contrast to the
screws for human use, screws for monkey skulls are shorter,
given the lower thickness of the macaque skull of around
3 mm. We redesigned the screws to be as short as possible,
i.e. with an overall length of 8 mm, 3 mm of which protrude
from the skull. The protruding section will be mechanically
stabilized by dental acrylic. The inner guiding lumen of the
bone screw is 6 mm long and has an inner diameter of
840±10 µm. The mismatch between inner screw and outer
10 mm
(a)
Ti ringZIF interface
16-ch
32-ch
(b) (c)
5 mm 15 mm
Chamber
Guiding screw/probeSkull replica
Probe tip Cap
Fig. 4. (a) Assembled 16- (top) and 32-channel probe (bottom) with Ti ring
and ZIF interface; (b) Hollow bone screws for probe alignment and fixation;
(c) skull replica with a mounted SEEG probe and protective chamber and
cap.
probe diameters will allow a maximum angular mismatch of
0.54◦in probe orientation. With a penetrating probe length of
32 mm this would result in a positional mismatch of 300 µm
at the probe tip. Not included in this mismatch calculation
is a possible malposition of the screw itself caused by the
surgical procedure.The correct probe position in depth is
ensured using a titanium (Ti) ring (thickness 1 mm, inner
diameter 0.9 mm, outer diameter 2.5 mm) adhesively fixed
to the probe using medical grade epoxy resin. The Ti ring
itself is fixed to the screw using bone cement. Fig. 4(c) shows
a macaque skull replica with a 16-channel probe inserted
through a guiding screw. The screw and the protruding ZIF
interface are mechanically protected by a chamber witch a
cap, shown in blue on Fig. 4(c). Both protective components
are designed with the focus on low space consumption in
order not to disturb the animals behaviour.
V. PROBE CHARACTERIZATION
A. Probe Geometry
The probe geometry was analyzed using a calibrated
optical microscope. In a fabrication run with 9 SEEG probes,
probe diameters of 794±6 µm were achieved.
B. Electrode Impedance Spectroscopy
The electrode impedance was characterized in 0.9% saline
solution using an electrochemical impedance analyzer (Com-
pactStat, Ivium Technologies, The Netherlands), in a 3-
electrode setup including a Pt counter electrode, a Ag/AgCl
reference electrode, and the electrode of a neural probe
applied as working electrode. Figs. 5(a) and (b) show abso-
lute impedance and phase values as a function of frequency
for macro- as well as micro-electrodes and their respective
standard deviations. The data were averaged over 15 macro-
10
2
10
3
10
4
10
5
10
6
-90
-80
-70
-60
-50
-40
-30
-20
-10
0
10
10
2
10
3
10
4
10
5
10
6
10
3
10
4
10
5
10
6
(a)
(b)
Frequency (Hz)
Frequency (Hz)
Impedance (Ω)Phase (degree)
Micro-electrodes
Macro-electrodes
Micro-electrodes
Macro-electrodes
Fig. 5. Frequency dependent (a) absolute value and (b) phase of the
impedance of macro- (n = 15) and micro-contacts (n = 30).
and 30 micro-electrodes of two 32- and one 16-channel
probe; impedance values at 1 kHz are 2150±483 Ωand
243±57 kΩfor the different electrode sizes, respectively.
C. Magnetic Resonance Imaging
To post-operatively determine the probe positioning in the
brain, magnetic resonance imaging is useful in particular in
view of functional brain imaging. However, MRI tends to be
obstructed by image distortion and signal voids caused by the
probe geometry and differences in the magnetic susceptibil-
ities of the applied materials. Large artifacts potentially lead
to a complete obstruction of magnetic resonance images. On
the other hand, small artifacts will hinder the identification
of tissue directly in contact with the probe. In order to
validate the new SEEG probes and their mounting screw,
MR images of the complete setup were taken inside a Bruker
Biospec 94/20 9.4 T MRI machine using Gd-doped water
as a contrast medium surrounding the sample in a ferrule
Fig. 6(a). Imaging was performed with a multi-spin echo
sequence (RARE, [10]) with TR = 2 s, TE = 28 ms, and a
voxel size 0.16 ×0.16 ×1 mm3. The representative image in
Fig. 6(b) shows only minor artifacts around the probe, which
do not interfere with the extraction of the probe position.
From the image, the diameter of the probe is determined
as 0.9 mm, which corresponds to a field-of-view loss of
0.05 mm on either side of the probe. As expected, the Ti
screw causes larger, yet localized, artifacts.
D. Probe Insertion Into Brain Phantom
Insertion tests were performed using a brain phantom
made of 0.6% agarose gel that exhibits mechanical char-
acteristics similar to real brain tissue [11]. Neither probe
buckling nor probe damage were observed during and after
multiple manual insertions of individual probes. Fig. 7(a)
shows a probe inserted into the brain phantom. Probe traces
were obtained by dipping probes into colored ink prior to
their insertion. An example of a trace is shown in Fig. 7(b).
Traces indicate a straight insertion path within the phantom.
2.5 mm
(a) (b)
Fig. 6. (a) SEEG probe in a ferrule filled with a contrast medium and (b)
MR image taken at 9.4 T.
(a) (b)
15 mm 5 mm
Fig. 7. (a) SEEG probe inserted into brain phatom and (b) traces of ink
after insertion.
VI. CONCLUSION
We have introduced a novel fabrication concept for SEEG
neural probes based on a PI layer sandwich. The design
and positioning of the electrode sites can easily be adapted
to the requirements for individual experiments. To keep the
connector space for the test on macaque monkeys as small
as possible, we produced 16- and 32-channel versions. In
the case of human use, we see no obstacle to extending
this to 64 or even 128 channels without the need for
integrated electronics to time-multiplex the different channels
in order to minimize the connector size. With a micro-
electrode impedance of 245 kΩat 1 kHz we expect that
it will be possible to record single unit activity from the
micro-contacts. Experiments in macaque monkeys will have
to confirm our concept regarding measurement quality and
biocompatibility. In the future, the electrode material can be
changed to iridium oxide in view of electrical stimulation of
neural tissue.
ACKNOWLEDGMENT
The authors would like to acknowledge technical discus-
sions and advice in PI process technology by Juan Ordonez,
Laboratory for Biomedical Microtechnology at IMTEK, and
discussions on the biological application with Leonardo
Fogassi and Fausto Caruana, Universit`
a degli studi di Parma.
REFERENCES
[1] J.M. Scarabin, Stereotaxy and Epilepsy Surgery. Paris: John Libbey
Eurotext, 2012.
[2] A. Palmini et al., “Focal neuronal migration disorders and intractable
partial epilepsy: results of surgical treatment,” Annals of Neurology,
vol. 30, no. 6, pp. 750-757, 1991.
[3] E. Toader et al., “Steering deep brain stimulation fields using a high
resolution electrode array,” in Engineering in Medicine and Biology
Society (EMBC), 2010 Annual Int. Conf. of the IEEE, Buenos Aires
2010, pp. 2061-2064.
[4] A. Mercanzini and P. Renaud, “Microfabricated neurostimulation
device,” European Patent Application EP2604313 A1, 2013.
[5] T. Stieglitz et al., “Micromachined, polyimide-based devices for flex-
ible neural interfaces,” J. Biomedical Microdevices, vol. 2, no. 4, pp.
283-294, 2000.
[6] J.S. Ordonez et al., “A 232-channel retinal vision prosthesis with
a miniaturized hermetic package,” in Engineering in Medicine and
Biology Society (EMBC), Annual Int. Conf. of the IEEE, San Diego
2012, pp. 2796- 2799.
[7] S. Kisban et al., “A novel assembly method for silicon-based neural
devices,” in World Congr. Med. Phys. and Biomed. Eng., 2009,
Munich, vol. 25/9, pp. 10710.
[8] P. Ruther et al., “Recent progress in neural probes using silicon MEMS
technology,” IEEJ Trans., vol. 5, pp. 50515, 2010.
[9] J.S. Ordonez et al., “Long-term adhesion studies of polyimide to
inorganic and metallic layers,” in MRS Proc., vol. 1466, 2012.
[10] J. Hennig et al., “RARE imaging: A fast imaging method for clinical
MR.,” Magn. Reson. Med., vol. 3, pp. 823833, 1986.
[11] C. Zhi-Jian et al., “A realistic brain tissue phantom for intraparenchy-
mal infusion studies,” J. Neurosurgery, vol. 101, pp. 314-322, 2004.
